Home

Heterogeneous MAS with Dynamic LLM Selection

Welcome to the wiki for the Heterogeneous Multi-Agent System with Dynamic LLM Selection project. This research prototype, developed at Ericsson Research, introduces a novel framework to overcome the limitations of homogeneous AI systems by leveraging a diverse set of Large Language Models (LLMs) through intelligent, dynamic selection.

This wiki provides detailed documentation on the project's architecture, setup, and usage. 📖 Getting Started

[[Installation]] - How to set up the project and its dependencies.

[[Quickstart Guide]] - A simple guide to run your first demo.

🧠 Understanding the Project

[[Core Concepts]] - Explanation of MAS, Heterogeneous LLMs, MCP, and the problem we solve.

[[Architecture Overview]] - A high-level view of the system's components and how they interact.

[[The Agentic AI Orchestrator]] - Deep dive into the brain of the system.

[[The MCP Server & Score Matrix]] - Details on the central knowledge base for LLM performance.

[[The Extended Benchmark Suite]] - Information on the test bench and how to use it for evaluation.

⚙️ For Developers

[[Configuration]] - How to configure the system, add new LLMs, and set priorities.

[[API Reference]] - Detailed API documentation.

[[Extending the Framework]] - How to add new domains, metrics, or algorithms.

🔍 More Information

[[Frequently Asked Questions]] - Common questions and answers.

[[Theory & Background]] - A deeper look at the research, including comparisons with Multi-Armed Bandits.

[[Future Work]] - Planned features and research directions.